WebXR version
Immersive Experience
Step inside the work. A guided spatial presentation brings Ben’s test architecture, agent workflows and engineering practice into view. Listen to the narration, follow the captions, and explore the structures as the story unfolds.
About 26 minutes, in nine chapters. Designed for a seated session, with controls within reach. Select an object to give it focus; the narration continues. The complete transcript is below.
The narration is unavailable. The show is paused. Retry the audio, or read the full narration below.
Read the full narrationThe 3D view is unavailable. The show is paused. Reload the view, or read the complete narration below.
Read the full narrationExplore this scene
Choose a structure to highlight it. You can also select it in the scene with a pointer or a headset ray. The voice keeps its place.
Read at your paceThe full narration
Everything the narrator says, in order. Ben’s own words are on the main page.
WelcomeBen McNulty
Welcome to the Immersive Experience. This is a guided tour of the career record of Ben McNulty, AI-Enhanced Senior Software Engineer, told in light. Everything you will see and hear comes from his website.
Ben’s focus: AI developer tooling, test architecture and reliable llm features.
Ben builds software across backend, frontend and testing, with a growing focus on how people and AI agents work together. He’s interested in new interfaces that make that collaboration useful and controllable, with the context, clear expectations and verification needed to get reliable work done.
Ben is looking for agent and developer tooling, test and evaluation infrastructure, applied AI engineering, or senior full-stack work on AI-native teams. He is based in Florida, USA, works remotely, and is open to conversations.
01 / Selected proofFollow the evidence.
Part one, selected proof. Follow the evidence. Three featured examples, each with the problem, Ben’s part and the key decision, then more write-ups on BenLive, his personal site. Each is labeled as current-role work, a personal project or first-person writing.
Cross-driver test architecture in C#. From his current role. The problem: A large VB.NET Selenium suite gates releases, and rewriting it in one go would pause regression coverage. Ben’s part: He architected the C# migration onto an adapter that runs Selenium or Playwright behind one interface. The key decision: The language move and the driver move stay separate: tests target the adapter, the driver is configuration, and a ported test passes on the original driver before it runs on the new one. Status: Port under way. The code is private; the sketch and the write-up show the architecture.
Resilience for a live AI feature. A personal project. The problem: Free-tier LLM providers fail in ordinary ways: 502s, sudden rate limits, models retired without notice. Ben’s part: He built the resilience layer behind his site’s public chat, in one Firebase Function. The key decision: An ordered model chain with a per-model circuit breaker, so a visitor gets an answer from a smaller model before an error from a bigger one; only allowlisted models run. Status: Live on his personal site. The backend code is not public.
AI Lab: tools for working with AI. A personal project. The problem: Make useful AI tools that people can direct and understand, across more than one isolated experiment. Ben’s part: He designs and maintains his personal AI Lab with AI coding assistants: Chat About Ben, Promptpad, Problem Solving and Agent Agenda. The key decision: Keep expectations and prompt contracts explicit, expose results for review, and give each tool controls suited to its job. Status: Public personal projects on BenLive. Not employer work.
Orgos: an agent-first organizational OS. A personal project. A personal agent-first OS Ben is building: agent identities, memory, files and flow sequencing live on disk as readable JSON and Markdown, with inference dispatched through an observable queue.
Version-controlled assistant rules. From his current role. The Rules, Profiles and Prompts Ben authored for his QA team, how edits to them are reviewed like code, and how new rules grow out of failures.
The whole delivery surface, directed and owned. A personal project. Ben designed the serverless backend, accessible vanilla-JS frontend, cross-browser tests, media pipeline and gated deploy, and he reviews and maintains them; AI coding assistants carried out the plans he approved.
Clear expectations for collaborative AI. A personal project. Versioned prompt contracts, sessions that can be rewound, and clear control states help people guide agents and inspect their work.
02 / Role fitRead the record your way.
Part two, role fit. Read the record your way. On the site, a visitor can choose the perspective closest to their role, and the career record highlights the parts it draws on.
There are six perspectives: Overview, AI tooling, Test infrastructure, Full-stack delivery, Frontend engineering and Ownership. Each changes what the record emphasizes, never the order of events.
Current role first, then earlier full-stack and frontend roles; personal AI projects are labeled as personal throughout. The record shows his roles in order, with the two independent periods grouped under one business.
03 / Career recordRoles, in order.
Part three, career record. Roles, in order. Personal AI projects are covered in the proof and capabilities sections.
QA Automation Engineer III. March 2022–present. He is the AI-assisted development-tooling point person for an eight-person automated regression testing team: he set up Kiro, the team’s current AI tool, and owns its supporting tooling, as he did for Amazon Q Developer, whose version-controlled Rules, Profiles and Prompts he authored. Architected the C# migration of a legacy VB.NET test suite onto a cross-driver adapter that runs Selenium or Playwright under the same test logic; the port is under way. Built and shipped an internal documentation and onboarding app that consolidates the team’s scattered docs into Markdown, with a web view built from the codebase. He drives weekly regression triage end to end, converting manual test paths into automated tests.
Independent Web Developer · McNulty Web Apps. June 2018–February 2022, and an earlier, separate period, 2012–2014. Carried responsive full-stack client projects from the first conversation through launch and support, translating ambiguous requests into practical frontend, backend and deployment decisions.
Frontend Engineer · Menuat. February 2016–May 2018. Shipped digital menu systems from concept to cloud-hosted production, translating brand requirements into responsive, accessible experiences on low-performance, TV-connected hardware. Covered project management in the interim, owning scoping, delivery and support across the client base until a permanent PM was hired.
Full-Stack Developer · Enix Networks. September 2012–August 2013. PHP page servers, contact forms and data handling for client sites.
04 / CapabilitiesWhat Ben brings to a team.
Part four, capabilities. What Ben brings to a team. Seven disciplines, each with practices, tools and evidence.
One of seven. AI-assisted engineering workflow. Ben works with agents by giving them relevant skills, context and clear expectations, then checking what they produce. His current personal work spans dots, Claude Duet and Muse; Claude Code and Codex are his primary direct coding interfaces. Earlier projects include Local Crew, a local-model orchestrator.
The practices: Give agents the relevant project context, skills and working agreements. Choose the interface and model for the work, with a clear handoff between agents. Make requests, constraints and expected results explicit before a run. Let agents carry work forward within agreed boundaries. Keep changes reviewable and check the result against the request. Record what changed, what was verified and what remains open.
Tools and methods: Claude Code, Codex, dots, Claude Duet, Muse, Ollama, Kiro, Amazon Q Developer and GitHub Copilot. The evidence, in his write-ups: “Models and Harnesses”, “A CLI Agent Harness Playbook”, “Designing past the AI template, prompt by prompt” and “Building Local Crew”.
Useful for teams that are adopting AI coding assistants and want reviewable diffs, explicit authority and verified handoffs rather than unbounded agent output. At work he set up Kiro, the QA team’s current AI tool, and owns its supporting tooling, as he did for Amazon Q Developer before it; that and the AI-assisted porting in the driver migration are QA-team scope. GitHub Copilot is his personal VS Code experience, going back to its beta. Everything else here is personal practice on his own projects and public site.
Two of seven. Context engineering and team AI rules. Ben turns scattered information into context people and agents can use. For an eight-person automated regression testing team, he set up Kiro and owns its supporting tooling, as he did for Amazon Q Developer. He codifies internal engineering standards and procedures in agent-usable documentation, and built a documentation app that brings team material into Markdown and a web view.
The practices: Rules, Profiles and Prompts reviewed in pull requests like code. A regression an assistant introduces becomes a new rule. One documentation path for new engineers and new assistant sessions. One canonical policy file that each assistant’s instruction file points to. Docs change in the same pass as the behavior they describe. Agent-facing guides such as llms.txt generated from one manifest.
Tools and methods: Amazon Q Developer Rules, Profiles & Prompts, Kiro, CLAUDE.md, AGENTS.md, Claude Code skills, Markdown and llms.txt. The evidence, in his write-ups: “Version-Controlled AI Governance for Engineering Teams”, “Context Engineering: The Skill That Makes AI Assistants Useful” and “Meta Programming the Workflow”.
Useful for teams that keep knowledge in people’s heads or scattered docs, and want new engineers and new assistant sessions to start from the same reviewed context. The Amazon Q Developer Rules, Profiles and Prompts, the Kiro tooling and the internal documentation app are QA-team work for his current employer. The canonical CLAUDE.md and AGENTS.md pattern, Claude Code skills and the llms.txt manifest come from his personal projects.
Three of seven. Verification and test architecture. Ben starts with a question: how could this passing result be false? At work, he architected a legacy suite’s move to C# on an adapter that runs the same tests through Selenium or Playwright. The port is under way. In his personal AI Lab, he audits test runners so interrupted batches cannot report success, and turns bugs he finds into automated checks.
The practices: Test runners audited so a killed batch cannot report green. Every hand-found bug becomes an automated check. Test logic never knows which browser driver runs underneath it. Ported tests keep running on the original driver until a switch is judged safe. Page objects expose intent, not locators and waits. Weekly regression triage, driven end to end.
Tools and methods: C#, Playwright, Selenium WebDriver, Cross-driver adapter, Page objects, axe-core and Firebase Emulator. The evidence, in his write-ups: “Cross-Driver Test Architecture”, “AI Lab Part 1: AI Chat”, “Shipping AI Features End to End as a Solo Engineer” and “A CLI Agent Harness Playbook”.
Useful for teams that carry a legacy suite, are changing drivers or frameworks, or are letting AI assistants touch code, and need a passing build to mean something. The adapter migration and regression triage belong to his current QA Automation Engineer III role. Runner audits, cross-browser suites and the bug-to-check rule come from his personal AI Lab.
Four of seven. Reliable, honest LLM features. Ben builds AI features with clear expectations, visible results and room for the person to direct the work. In his personal AI Lab, chat replies name the model that answered; versioned prompt contracts keep behavior explicit; and an allowlisted model chain with a circuit breaker is designed to return an answer from a smaller model rather than fail.
The practices: Clear expectations and versioned prompt contracts that make agent behavior reviewable. Chat replies labeled with the model that actually answered. Allowlisted models, a fallback chain and a per-model circuit breaker. System prompts kept server-side behind versioned prompt keys. Separate server and client retry policies, each traced to a real failure. Interfaces that help people guide the work, inspect results and build trust.
Tools and methods: OpenRouter, Ollama and Firebase Functions. The evidence, in his write-ups: “AI Lab Part 1: AI Chat”, “AI Lab Part 3: Problem Solver”, “AI Lab Part 4: Agent Agenda” and “Production Resilience Patterns for AI-Powered Features”.
Useful for teams that put language models in front of users, where cost control, honest attribution, privacy boundaries and graceful degradation have to be designed in from the start. Personal AI Lab work on his own public site.
Five of seven. Full-stack product delivery. Ben takes web products from the first request through deployment and support, in TypeScript, JavaScript, React, Next.js, Node.js, C# and PHP. As an independent developer he carried client projects from the first conversation to support; earlier he built PHP client sites at Enix Networks. His recent personal work includes a serverless AI backend that he designed, built with AI coding assistants and maintains as its only engineer.
The practices: Serverless backends on Firebase Functions and Firestore. Rate limiting keyed on an HMAC of the visitor’s IP, never the raw address. Personal-site deploys gated by model probes and post-deploy smoke checks. One shared module where a configuration had been forked. Client delivery from first conversation through deployment and support. PHP page servers, contact forms and data handling.
Tools and methods: TypeScript, JavaScript, React, Next.js, Node.js, C#, PHP, REST APIs, SQL/NoSQL and Firebase. The evidence, in his write-ups: “Shipping AI Features End to End as a Solo Engineer” and “Production Resilience Patterns for AI-Powered Features”.
Useful for teams that need one engineer to carry a feature across backend, frontend, tests and deploy, and to explain the tradeoffs between those layers. The Firebase and serverless work is his personal AI Lab. Independent Web Developer, June 2018–February 2022. Full-Stack Developer at Enix Networks, September 2012–August 2013.
Six of seven. Interface, motion and spatial UI. Ben builds responsive, accessible interfaces, from advanced CSS and design systems to spatial UI. At Menuat he shipped animated digital menu systems on low-performance, TV-connected hardware. For his own site he directed a Codex-assisted rebuild of a Three.js and WebXR gallery, with hand-tracking grabs, intent-safe teleport and an AR miniature, and kept a route that needs no 3D.
The practices: Animated digital menu systems designed for constrained, TV-connected hardware. Brand requirements translated into responsive on-screen experiences. Hand-tracking grabs, intent-safe teleport and a hand-held AR view. Design systems built on oklch, clamp() type and container queries. Contrast checked in both themes; focus kept keyboard-safe. A stable, readable reduced-motion version of every effect.
Tools and methods: CSS, Three.js, WebXR, ARIA, Container queries, oklch and cannon-es. The evidence, in his write-ups: “Rebuilding my WebXR portfolio, one headset test at a time” and “Designing past the AI template, prompt by prompt”.
Useful for teams that build visual or immersive products that must stay readable, accessible and smooth on real devices, not only in a desktop browser. Frontend Engineer at Menuat, February 2016–May 2018. The WebXR gallery and site visuals are personal, AI-assisted projects that he directed and verified.
Seven of seven. Communication and handoffs. Ben connects the people and disciplines a project needs: users, design, frontend, backend and testing. At Menuat he was the frontend intermediary between users, design and backend. He makes expectations clear, writes down decisions and leaves handoffs someone else can pick up; those habits also shape how he collaborates with agents.
The practices: Every run ends with a handoff: what changed, what was verified, what was not. Decision logs record why, so settled questions stay settled. Handoffs designed so a qualified stranger could continue safely. Briefs that ask for plain, action-first on-screen copy. Candid retrospectives, including what did not work. Frontend intermediary between users, design and backend at Menuat.
Tools and methods: Decision logs, Handoff notes, Markdown docs and Pull-request summaries. The evidence, in his write-ups: “A CLI Agent Harness Playbook”, “Meta Programming the Workflow” and “AI Lab Part 5: Iterative Engineering with AI Assistants”.
Useful for teams that work remotely and in writing, and need someone who leaves work in a state that other people, or assistants, can pick up. Frontend Engineer at Menuat, February 2016–May 2018. The rest is practice shown in his own writing and projects.
05 / Where the skills meetThe value is in the combinations.
Part five, where the skills meet. The value is in the combinations. Hiring for one skill rarely fills a role. Each pairing joins two disciplines in one project or article.
Test architecture and AI workflow. Assistants work inside the test contract. Alongside the driver migration, Ben codifies the team’s internal engineering standards and procedures in documentation agents can use. The versioned Rules and Profiles make the adapter contract and expected checks explicit; people own the architecture and release decisions.
Context & team rules and AI workflow. Harnesses judged on context hygiene. Ben rates a coding harness first on context hygiene: it has to read the project’s rules file on every session without anyone pasting it in.
Context & team rules and Test architecture. Rules that require tests. The QA team’s repository-owned AI rules require every behavior change to arrive with its matching test update, and changes to those rules are reviewed in pull requests.
AI product reliability and Test architecture. User-gated AI that stays testable. Each model call waits for a click, so end-to-end tests can mock the model and assert the control state at every step. Those tests check the controls; the model’s variable answers are a separate question.
AI product reliability and Full-stack delivery. Abuse protection that respects privacy. The public chat’s rate limiter keys on an HMAC of the visitor’s IP under a secret salt, so raw addresses are never stored or logged.
AI product reliability and AI workflow. Model choice that plans for failure. The rule he uses to pick models, the cheapest one that clears the bar for the task, with a fallback, reproduced locally before shipping, is the rule the public chat runs on.
AI product reliability and Interface & spatial. A user-triggered editor, still under test. Promptpad shipped a user-triggered inference loop, a revision model and a polished editor together, and the interface stayed testable.
Interface & spatial and Test architecture. Automated checks for spatial interaction. Automated checks cover the WebXR gallery’s hand gestures, teleporting and AR entry and exit; headset sessions then look for what the tests missed.
AI workflow and Interface & spatial. Directing AI design, judged on device. For the October polish pass on his site, Ben directed the work and judged each change on his phone; Claude did the design and engineering.
Full-stack delivery and AI workflow. Assistants run the loop; he owns the outcome. Across his AI Lab’s delivery surface, assistants carry context and execute plans he has approved. Ben decides what to build, sets the constraints, reviews the plan and the diff, and owns the outcome.
Full-stack delivery and Test architecture. Deploys gated by probes and smoke checks. Before each release of his personal site, the deploy path probes every live model with a tiny request; afterward a smoke script checks status codes and headers.
Context & team rules and AI product reliability. System prompts stay on the server. AI Lab apps send a prompt key, never a prompt, so system prompts are not shipped to the browser and a visitor can choose only from allowed keys. That narrows what a visitor controls; it is not a guarantee against prompt injection.
Collaboration and Interface & spatial. Plain language on screen. In the brief for his site’s launch reel he asked for implementation jargon to be replaced with plain language that tells the viewer what they can do, then reviewed the on-screen copy Claude rewrote.
Collaboration and Test architecture. Handoffs that say what was not verified. A run is not finished until its handoff lists the files changed, what was verified and what was not.
Collaboration and Context & team rules. Decisions that stay decided. A decision log records why each choice was made, so the next person or assistant session starts from settled decisions.
Collaboration and AI workflow. Handoffs a stranger could continue. Ben designs each agent run to end in a handoff that a qualified stranger, person or assistant, could pick up safely.
06 / How Ben worksWorking principles.
Part six, how Ben works. Working principles. The principles behind Ben’s work, each argued in his write-ups.
Ask how a green result could be false. A runner that treats a killed process as a pass looks green right up until it matters, so every passing result has to be honest about how it could be wrong.
Harness before model. He chooses the working loop first, judging its context, change control, verification and handoff, and only then debates which model runs inside it.
Context decides the result. Model quality sets a ceiling. The context an assistant starts with decides how close the work gets to it.
Write the team’s AI rules down. Every team using assistants already has rules. They belong in a reviewed file, not in each engineer’s habits.
Spend inference only on request. His AI Lab apps do not spend a generation the user did not ask for; the person deciding what happens next is the interface, not a limitation on the AI.
Degraded, not down. A visitor should get an answer from a smaller model before they get an error from a bigger one.
A stale document is a bug. Assistants execute against documentation the way they execute against code, so docs change in the same pass as the behavior they describe.
Team fitTeams where this combination fits
Teams where this combination fits.
Engineering teams formalizing how they work with AI assistants. Assistant rules, prompts and review practice versioned in the repository, with harness and model choices made per task. For an eight-person automated regression testing team, Ben is the AI-assisted development-tooling point person and authored its Rules, Profiles and Prompts.
Product teams shipping LLM features to real users. Allowlisted models, fallbacks, honest attribution and privacy boundaries have to be designed together with a tested full-stack implementation.
Test-infrastructure teams modernizing a legacy suite. Frameworks or drivers are changing under existing coverage, release gates have to hold during the change, and assistants need written rules before they touch tests.
Product teams that need a senior full-stack engineer. One engineer carries a feature across backend, frontend, tests and deploy, and can explain the tradeoffs between those layers.
07 / Next conversationLet’s talk about the work.
Part seven, next conversation. Ben’s invitation: let’s talk about the work. Ben is looking for agent and developer tooling, test and evaluation infrastructure, applied AI engineering, or senior full-stack work on AI-native teams. He is based in Florida, USA, works remotely, and is open to conversations.
You can email Ben at benmcnulty@live.com, or find him on LinkedIn and GitHub. Thank you for joining the Immersive Experience.